# loading libraries
library(dplyr)
library(ggplot2)
library(ggpubr)
library(tidyverse)
library(plotly)
library(httpuv)
library(lubridate)
library(RColorBrewer)
library(scales)
library(stringr)
library(readr)
library(gridExtra)
library(maps)

# modifying chart size
options(repr.plot.width=5, repr.plot.height=3)

#upload excel file
flights0 <- read_csv("flight_data/bos_flights_only.csv")
airline <- read_csv("flight_data/airlines.csv")
airports <- read_csv("flight_data/airports.csv")

flights <- flights0 %>% 
  left_join(airports, by = c("DESTINATION_AIRPORT" = "IATA_CODE")) %>% 
  left_join(airline, by = c("AIRLINE" = "IATA_CODE"))

Project Overview: Boston Logan airport is the largest airport in New England with over 100,000 domestic flights annually. While the airport does offer exceptional ocean views and dining options, ie Dunkin Donuts, delayed and canceled flights can create financial and emotional burdens for passengers like increased trip costs and missed work or vacation days. To maximize our trips and avoid flight delays and/or cancellations out of Boston Logan we will explore which airlines, departure times, and destinations experience the fewest disruptions. We will also explore if a flight is delayed what factors contribute to whether or not the flight will ultimately arrive on time or late at the destination airport.

Data: To explore our question above we downloaded 3 datasets related to domestic flights in the United States in 2015. The full datasets can be found here: https://www.kaggle.com/usdot/flight-delays/version/1#flights.csv

Dataset 1 - Flights: The flights dataset contains a row for every domestic flight in the United States in 2015. The original file contained 5.8 million rows and was 1.1 GB. After uploading the file into Big Query, we ran a query to capture only flights that departed out of Boston to increase query processing time and reduce the amount of money we needed to pay per query. Our updated dataset contained 107847 rows and was 20 MB. We used the following columns for our analysis. Note: There was no data available for flights out of Boston in October in this dataset.

YEAR - year of the flight; dbl class; only includes data from 2015

MONTH - month of the flight; dbl class

DAY - day of the flight; a dbl class

AIRLINE - code of the airline; chr class

FLIGHT_NUMBER - flight number; dbl class

ORIGIN_AIRPORT - the departing airport; chr class; identical to the IATA_CODE column from the airport table

DISTANCE - the distance in miles between the origin airport and destination airport; dbl class

DESTINATION_AIRPORT - the destination airport; chr class; identical to the IATA_CODE column from the airport table

DEPARTURE_DELAY - the total departure delay time in minutes; dbl class

ARRIVAL_DELAY - the total arrival delay time in minutes; dbl class

AIR_SYSTEM_DELAY - the departure delay in minutes caused by an air system issue (non-extreme weather conditions, airport operations, heavy traffic volume, air traffic control, etc.); dbl class

SECURITY_DELAY - the delay in minutes caused by a security issue (evacuation of a terminal/concourse, re-boarding of aircraft because of a security breach, inoperative screening equipment and/or long lines in excess of 29 minutes at screening areas); dbl class

AIRLINE_DELAY - the delay in minutes caused by an airline issue (technological, maintenance and fueling problems with the airline); dbl class

LATE_AIRCRAFT_DELAY - the delay in minutes caused by the late arrival of the aircraft to be used for the flight from a previous flight and congestion in air traffic; dbl class

WEATHER_DELAY - the delay in minutes caused by extreme or hazardous weather conditions that are forecasted or manifest themselves on point of departure, enroute, or on point of arrival; dbl class

Dataset 2 - Airlines: The airline dataset contains two columns, the name of the 14 airlines with domestic flights within the U.S. in 2015 and their airline codes and 15 rows. Because the flights dataset did not contain full airline names, we had to join the airlines table to the flights table by airline code to reference airline names. The following columns were availble in the airlines dataset:

Airline_name - the full name of the airline; a nullable string

IATA_code - the associated code with the airline; a nullable string

Dataset 3 - Airports: The airports table contains seven columns and 322 rows. Because the flights dataset did not contain the full airport name or destination city, we used the IATA_CODE column to join the flights table and capture this information. The following columns were available in the flights dataset:

IATA_code - the associated code with the airport; a nullable string

Airport - full name of the airport; a nullable string

City - name of the city the airport is in; a nullable string

State - name of the state the airport is in; a nullable string

Country - name of the country the airport is in; a nullable string

Below is a sample of the raw data of the files:

head(flights)
head(airline)
head(airports)

Data Exploration

Question: How many airports can you fly to from Boston and what were the most popular airports to fly to?

In 2015, Boston Logan had flights to 62 unique airports in the United States. Ronald Reagan Airport in Arlington, LaGuardia in New York, and Chicago O’Hare were the most popular destinations in 2015 with each airline receiving over 6,500 flights from Boston.

flights %>% 
  count(AIRPORT, CITY, sort = TRUE)
flights %>% count(DESTINATION_AIRPORT, sort = TRUE) %>%
  arrange(desc(n)) %>%
  ggplot() +
  geom_col(aes(x = reorder(DESTINATION_AIRPORT,-n), y = n),fill = "#7570B3") +
  theme(axis.text.x=element_text(angle=90, hjust=1)) +
  labs(title = "Destination Airports By Flight Volume", x = "destination airport", y = "count of flights")

Question: How many flights are out of Boston each month?

Throughout the year there were between 8,300 and 10,800 flights out of Boston each month. January had the least amount of flights, with 8,837 while July had the most flights with 10,837. In general, the winter months were the least busy while the summer months were the busiest.

flights %>% count(MONTH, sort = TRUE) %>%
  arrange(desc(n)) %>%
  ggplot() +
  geom_col(aes(x = factor(MONTH), y = n),fill = "#E6AB02") +
  theme(axis.text.x=element_text(angle=90, hjust=1)) +
  labs(title = "Flights By Month", x = "month", y = "count of flights")

Question: How many flights departed from Boston on each day of the week?

Throughout the year, there were more flights out of Boston on weekdays compared to weekends, which is due to the notion that business trips are during the week and people who want to spend their weekend out of the city will usually fly Friday after work/school. Most flights out of Boston are on Mondays and the least are on Saturdays. Tuesdays, Wednesdays, and Thursdays have about the same number of flights departing Boston.

ggplotly(flights %>%
  mutate(date = make_datetime(YEAR, MONTH, DAY),
        weekday = wday(date, label = TRUE)) %>%
  group_by(weekday) %>%
  summarize(count = n(),  # counting the number of flights (per weekday)
            delay = mean(ARRIVAL_DELAY, na.rm = TRUE)  # average delay (per weekday)
           ) %>% 
  ggplot(aes(weekday, count)) + 
    geom_bar(stat="identity", width = 0.5, fill = "#7570B3") + 
    labs(title="Number Of Flights Per Day Of Week",
        x = "day of week",
        y = "count") +
    theme_linedraw() +
    theme(axis.text.x = element_text(angle=65, vjust=0.6)))

Question: How many flights departed late out of Boston Logan in 2015?

In 2015 39,093, or 37% of flights departed late out of Boston.

delay_count <- flights %>%
 transmute(DESTINATION_AIRPORT, DEPARTURE_DELAY, DELAYED = DEPARTURE_DELAY, NOT_DELAYED = DEPARTURE_DELAY)
delayed_istrue <- which(delay_count$DEPARTURE_DELAY > 0)
delayed_isnottrue <- which(delay_count$DEPARTURE_DELAY <= 0)
delay_count$DELAYED[delayed_istrue] <- 1
delay_count$DELAYED[delayed_isnottrue] <- 0
delay_count$NOT_DELAYED[delayed_istrue] <- 0
delay_count$NOT_DELAYED[delayed_isnottrue] <- 1
delay_count %>%
 select(DESTINATION_AIRPORT, DELAYED, NOT_DELAYED) %>%
 filter(!is.na(DELAYED), !is.na(NOT_DELAYED)) %>%
 summarize(total_delays = sum(DELAYED), total_ontimes = sum(NOT_DELAYED)) %>%
 mutate(percent_delayed = total_delays/(total_delays + total_ontimes))

In general, in 2015 there were more on-time flights than delayed flights on any given day out of Boston Logan. Typically canceled flights made up a very small portion of total flights, with the exception being 1/27, 2/2, 2/9, and 2/15 where we can see the majority of flights were canceled.

flights %>%
 select(YEAR,MONTH,DAY,DEPARTURE_DELAY,CANCELLED) %>%
 mutate(date = make_date(YEAR,MONTH,DAY), delay10 = ifelse(DEPARTURE_DELAY > 0, 1, 0), notdelay10 = ifelse(DEPARTURE_DELAY <= 0, 1, 0)) %>%
 group_by(date) %>%
 summarize(delayed = sum(delay10, na.rm = TRUE), notdelayed = sum(notdelay10, na.rm = TRUE), cancelled = sum(CANCELLED, na.rm = TRUE)) -> flightsByDay
f <- ggplot(data = flightsByDay) +
 geom_line(aes(x = date, y = delayed), color = "#E6AB02") +
 geom_line(aes(x = date, y = notdelayed), color = "#66A61E") +
 geom_line(aes(x = date, y = cancelled), color = "#E7298A") +
 labs(title = "On-Time, Delayed, and Cancelled Flights by Day", x = "date", y = "count of flights")
ggplotly(f)

Question: What is the trend line for the daily average departure delay throughout the year?

The month of February in 2015 had the highest daily average delay in minutes with February 2nd standing out for having average departure delays of 144 minutes. February of 2015 was recorded as the snowiest month on record in Boston since 1891 with almost 65 inches of snow which likely led to an increase in flight delays. Another day that stood out was December 29th, for having the highest average delay of 154 minutes, this was likely due to the mix of a high volume of flights returning from Christmas, departing to celebrate New Years, and flight delays due to snow.

flight_count <- flights %>%
  mutate(date = make_datetime(YEAR, MONTH, DAY)) %>%
  group_by(date) %>%
  summarize(count = n(),  # counting the number of flights/day
            delay = mean(DEPARTURE_DELAY, na.rm = TRUE)  # average delay/day
           )
  
p <- ggplot(flight_count, aes(date, delay)) +
  geom_line(color="#E6AB02") +
  labs(title = "Daily Average Departure Delay Out Of Boston", y = "average delay (min)") +
  theme_classic()

ggplotly(p)
#by month 
bosFlights <- read_csv("flight_data/bos_flights_only.csv")
bosFlights %>% 
  mutate(monthFac = as.factor(MONTH)) %>% 
  group_by(monthFac) %>% 
  summarize(count = n(),
            delay = mean(DEPARTURE_DELAY, na.rm = TRUE)) -> delayByMonth

delayByMonth
ggplot(delayByMonth, aes(x = monthFac, y = delay)) +
  geom_col(fill="#E6AB02") + 
  labs(title = "Monthly Average Departure Delay", y = "average delay (min)", x = "month")

Question: Which destination has the highest proportion of departure delay out of all the destinations from Boston?

The chart below shows the top 7 destinations with the highest percentage of departure delays out of Boston Logan. Detroit Metro Airport appears to have a significantly higher percent of delay in comparison to all the other destinations out of Boston.

z <- flights %>%
 group_by(DESTINATION_AIRPORT) %>% 
 summarise(avgDelay=mean(DEPARTURE_DELAY,na.rm=T)) %>%
 filter(avgDelay>10) %>%
 select(avgDelay,DESTINATION_AIRPORT) %>%
 arrange(desc(avgDelay)) %>%
 ggplot(aes(DESTINATION_AIRPORT,avgDelay))+
 geom_bar(stat="identity",fill="#E6AB02")+
 geom_hline(yintercept=20,color="#E7298A",size=1)+
 labs(x="destination airport",
        y="average delay",
        title="Average Delay by Destination") +
     theme(axis.text.x=element_text(angle=90, hjust=1))

flights %>%
 filter(!is.na(DEPARTURE_DELAY),DEPARTURE_DELAY>0) %>%
 mutate(percDealy=DEPARTURE_DELAY/n(),na.rm=T) %>%
 filter(percDealy>0.02) %>%
        select(percDealy,DESTINATION_AIRPORT) %>%
          arrange(desc(percDealy)) %>%
ggplot(aes(DESTINATION_AIRPORT,percDealy))+
geom_bar(stat = "identity",aes(fill=DESTINATION_AIRPORT))+
coord_polar()+
labs(x="destination airport",
      y="percentage delay",
      title="Percentage Delay by Destination")

Question: What cities experienced the longest average arrival delays?

Akron experienced the longest arrival delays of any US city with an average arrival delay of 26 minutes followed by Sacramento with a delay of 19 minutes and San Francisco at 15 minutes.

flights %>%
  group_by(DESTINATION_AIRPORT) %>%
  summarise(avg_dest_delay = mean(DEPARTURE_DELAY, na.rm = TRUE)) %>%
  inner_join(airports, by = c("DESTINATION_AIRPORT" = "IATA_CODE")) %>%
  filter(LONGITUDE < 0 & between(LATITUDE, 25, 50))  -> delayByState # limiting the map to US mainland boarders    for clarity  # only destinations with delays more than one hour
 
delay_buckets <- cut(x = delayByState$avg_dest_delay, breaks = c(0,5,10,15,20,25,30))
levels(delay_buckets) <- c("<5", "5-10", "10-15", "15-20", "20-25", "25-30")

delayByState$delay_buckets <- delay_buckets
 
  ggplot(data = delayByState, aes(y = LATITUDE, x = LONGITUDE, color = delay_buckets, size = delay_buckets)) +
  borders("state") +
  geom_point(alpha = .5) +
  coord_quickmap() +
  scale_color_brewer(palette = "Dark2") +
  theme(panel.background = element_blank())

delayByState %>% 
  select(CITY, avg_dest_delay) %>% 
  arrange(desc(avg_dest_delay)) %>% 
  head(5)

Question Which airport had the most flights canceled?

LGA, LaGuardia, had the most flights canceled to it with almost 6% of flights departing for LGA ultimately being canceled.

flights %>%
  group_by(DESTINATION_AIRPORT) %>%
  summarise(percent_canceled = sum(is.na(DEPARTURE_DELAY))/n() * 100, n = n()) %>%
  filter(n >= 5) %>% 
  inner_join(airports, by = c("DESTINATION_AIRPORT" = "IATA_CODE")) %>%
  filter(LONGITUDE < 0 & between(LATITUDE, 25, 50)) %>%
  ggplot(aes(y = LATITUDE, x = LONGITUDE, size = percent_canceled)) +
  borders("state") + 
  geom_point(alpha = .5, color = "#E7298A") +
  coord_quickmap() +
  theme(panel.background = element_blank())

flights %>%
  group_by(DESTINATION_AIRPORT) %>%
  summarise(percent_canceled = sum(is.na(DEPARTURE_DELAY))/n() * 100, n = n()) %>%
  filter(n >= 5) %>% 
  select(DESTINATION_AIRPORT,percent_canceled) %>% 
  arrange(desc(percent_canceled))

Question: Does departing late guarantee that your flight will land late?

Even if your flight departs late it is possible that your flight will still land on time at the destination airport. About 12,000 flights that departed late out of Boston in 2015 landed on time.

bos_flights_delay_expand <- flights
yes <- which(bos_flights_delay_expand$DEPARTURE_DELAY > 0)
bos_flights_delay_expand$DEPARTURE_DELAY[yes] <- "yes"
no <- which(bos_flights_delay_expand$DEPARTURE_DELAY <= 0)
bos_flights_delay_expand$DEPARTURE_DELAY[no] <- "no"

bos_flights_delay_expand <- select(bos_flights_delay_expand,DEPARTURE_DELAY, ARRIVAL_DELAY, DISTANCE, AIR_TIME, DESTINATION_AIRPORT)

yes <- which(bos_flights_delay_expand$ARRIVAL_DELAY > 0)
bos_flights_delay_expand$ARRIVAL_DELAY[yes] <- "yes"
no <- which(bos_flights_delay_expand$ARRIVAL_DELAY <= 0)
bos_flights_delay_expand$ARRIVAL_DELAY[no] <- "no"

delayed_flights_only <- filter(bos_flights_delay_expand, DEPARTURE_DELAY == "yes")  

#chart one - how many flights that depart late also arrive late
ggplot(data = na.omit(delayed_flights_only), aes(x = ARRIVAL_DELAY)) +
  geom_bar(fill = "#D95F02") +
  labs(title = "Do All Flights That Depart Late Arrive Late?", x = "flight arrived late", y = "count of flights")

Question What impacts on whether or not a delayed flight will land on time?

In the chart below we can see that the longer the departure delay is in minutes, the less likely a flight is to land on time at the destination airport. For flights that departed less than 10 minutes late from Boston in 2015, almost 60% still landed on time at their destination airport. For flights that departed between 10 and 20 minutes late, over a quarter landed on time. However, 100% of flights that departed an hour or later also landed late. As we saw earlier, more than 1/3 of flights departed late out of Logan in 2015. Because of the likeliness of departing late airlines likely pad their flight times so that they are able to appear on schedule.

distance_buckets <- cut(x = delayed_flights_only$DISTANCE, breaks = c(0,500,1000,1500,2000,2500,3000))
levels(distance_buckets) <- c("<500","501 - 1000", "1001 - 1500", "1501 - 2000", "2001 - 2500","2501 - 3000",">3000")
delayed_flights_only$distance_buckets <- distance_buckets

bos_flights_delay_expand$dep_delay_minutes <- flights$DEPARTURE_DELAY
bos_flights_delay_expand$arr_delay_minutes <- flights$ARRIVAL_DELAY

summary(na.omit(flights$DEPARTURE_DELAY))
##     Min.  1st Qu.   Median     Mean  3rd Qu.     Max. 
##  -35.000   -5.000   -2.000    9.606    7.000 1190.000
bos_flights_delay_lt60 <- filter(bos_flights_delay_expand, dep_delay_minutes > 0 & dep_delay_minutes <= 60)  

dep_delay_buckets <- cut(x = bos_flights_delay_lt60$dep_delay_minutes, breaks = c(0,10,20,30,40,50,60))
levels(dep_delay_buckets) <- c("<10", "11 - 20", "21 - 30", "31 - 40","41 - 50","51 - 60")
bos_flights_delay_lt60$dep_delay_buckets <- dep_delay_buckets

ggplot(data = na.omit(bos_flights_delay_lt60)) +
  geom_bar(mapping = aes(x = dep_delay_buckets, fill = ARRIVAL_DELAY), position = "fill") +
  labs(title = "Arrival Time Of Delayed Flights By Departure Delay (min)", x = "departure delay (min)", y = "percent of on-time flights") +
  scale_fill_brewer(palette = "Dark2")

Question: Are flight times for longer distance flights padded more than for shorter flights?

As the distance of flights increases, there is a slightly higher chance that the flight will land on-time despite a departure delay. However, no flights that had a departure delay of over 50 minutes were able to overcome landing late no matter the flight distance.

distance_buckets_2 <- cut(x = bos_flights_delay_lt60$DISTANCE, breaks = c(0,500,1000,1500,2000,2500,3000))
levels(distance_buckets_2) <- c("<500","501 - 1000", "1001 - 1500", "1501 - 2000", "2001 - 2500","2501 - 3000",">3000")
bos_flights_delay_lt60$distance_buckets <- distance_buckets_2

ggplot(data = na.omit(bos_flights_delay_lt60)) +
  geom_bar(mapping = aes(x = dep_delay_buckets, fill = ARRIVAL_DELAY), position = "fill")+
  facet_wrap(~ distance_buckets) +
  scale_fill_brewer(palette = "Dark2") +
  labs(title = "Arrival Time Of Delayed Flights By Flight Distance And Delay Minutes", x = "departure delay (min)", y = "percent of on-time flights") +
  theme(axis.text.x=element_text(angle=90, hjust=1))

Question: What is the relationship between average distance and the average arrival delay of flights from Boston to another domestic airport?

The relationship between average distance and arrival delay based on all delayed flights appears to be split in different directions based on a cut-off distance with flights up to 1000 miles having a steeper negative relationship with arrival delay compared to the distance between 1000 and 2000 miles. However, flights that fly more than 2000 miles have a positive relationship with arrival delay that explains arrival delay increases for flights that have destinations over 200 miles from Boston.

p <- ggplot(data = delay, mapping = aes(x = distance, y = delay)) +
  geom_point(aes(size = count, color = DESTINATION_AIRPORT), alpha = 1/3) +
  geom_smooth(se = FALSE, color = "#E7298A") +
  labs(title = "Average Distance (mi) To Destination vs. Average Arrival Delay (min)",
      caption = "The circle size shows the number of flights to that destination.")

ggplotly(p + theme(legend.position = "none") + scale_fill_brewer(palette = "Dark2"))

Question Which airline pads their flight times the most?

Of the top four most popular airlines, it appears that all four airlines pad their flight times. If a flight departs on time we can see in general that it will land early, also the longer a flight is the more likely it is to land early if it departs on time. The pad time amongst airlines seems to be fairly similar, but we can see in the chart below that the estimated arrival line for United flights that depart on time is completely in the negative, meaning in general, if a United flight departs on time it will most likely land early. In contrast, American Airlines seem to pad their flights the least.

on_time_flights <- filter(flights, DEPARTURE_DELAY == 0)
on_time_flights_top_airlines <- filter(on_time_flights, AIRLINE == c("AA","B6","DL","UA"))

#ontime flights by airline
ggplot(data = on_time_flights_top_airlines, aes(x = DISTANCE, y = ARRIVAL_DELAY, color = AIRLINE.y)) +
  geom_jitter(alpha = .5, color = "#666666") +
  geom_smooth(method = "glm", se = FALSE, size = 1.5 ) + 
  scale_colour_brewer(palette = "Dark2") +
  labs(title = "Flight Time Padding By Airline", x = "flight distance", y = "arrival delay", color = "airline")

Question Which type of delay is the worst?

The dataset includes 5 major types of weather delays. The most common delay type is an air system delay with 14,179 flights impacted by this delay type at Logan in 2015. The average air system delay was 28 minutes long. The second most common delay type was airline delay which impacted 10,934 flights with an average delay time of 29 minutes. The third and fourth most common delay types, late inbound aircraft and weather delay, while being less common, had a longer average delay time of 43 and 44 minutes respectively.

air_system_delay <- filter(flights, AIR_SYSTEM_DELAY > 0, AIR_SYSTEM_DELAY < 300)
weather_delay <- filter(flights, WEATHER_DELAY > 0, WEATHER_DELAY < 300)
late_aircarft_delay <- filter(flights, LATE_AIRCRAFT_DELAY > 0, LATE_AIRCRAFT_DELAY < 300)
security_delay <- filter(flights, SECURITY_DELAY > 0, SECURITY_DELAY < 300)
airline_delay <- filter(flights, AIRLINE_DELAY > 0, AIRLINE_DELAY < 300)

air_system_delay <- filter(flights, AIR_SYSTEM_DELAY > 0, AIR_SYSTEM_DELAY < 300)
weather_delay <- filter(flights, WEATHER_DELAY > 0, WEATHER_DELAY < 300)
late_aircarft_delay <- filter(flights, LATE_AIRCRAFT_DELAY > 0, LATE_AIRCRAFT_DELAY < 300)
security_delay <- filter(flights, SECURITY_DELAY > 0, SECURITY_DELAY < 300)
airline_delay <- filter(flights, AIRLINE_DELAY > 0, AIRLINE_DELAY < 300)

airsystemPlot <- ggplot() +
  geom_histogram(data = air_system_delay, aes(x = AIR_SYSTEM_DELAY),binwidth = 5, fill = "#E7298A", color = "white") +
  ylim(0, 2250) + xlim(0,200)
airlinePlot <- ggplot() +
  geom_histogram(data = airline_delay, aes(x = AIRLINE_DELAY), binwidth = 5, fill = "#66A61E", alpha = .5, color = "white") +
  ylim(0, 2250) + xlim(0,200)
lateAircraftPlot <- ggplot() +
  geom_histogram(data = late_aircarft_delay, aes(x = LATE_AIRCRAFT_DELAY),binwidth = 5, fill = "#E6AB02", color = "white") +
  ylim(0, 2250) + xlim(0,200)
weatherDelayPlot <- ggplot() +
  geom_histogram(data = weather_delay, aes(x = WEATHER_DELAY), binwidth = 5, fill = "#7570B3", color = "white") +
  ylim(0, 2250) + xlim(0,200)
securityDelayPlot <- ggplot() +
  geom_histogram(data = security_delay, aes(x = SECURITY_DELAY),binwidth = 5, fill = "#666666", color = "white") +
  ylim(0, 2250) + xlim(0,200)

grid.arrange(airsystemPlot, airlinePlot, lateAircraftPlot, weatherDelayPlot, securityDelayPlot, nrow = 3)

summary(air_system_delay$AIR_SYSTEM_DELAY)
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##    1.00   10.00   19.00   28.77   33.00  299.00
nrow(air_system_delay)
## [1] 14179
summary(airline_delay$AIRLINE_DELAY)
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##    1.00    7.00   15.00   29.01   33.00  298.00
nrow(airline_delay)
## [1] 10934
summary(late_aircarft_delay$LATE_AIRCRAFT_DELAY)
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##    1.00   14.00   28.00   43.34   56.00  298.00
nrow(late_aircarft_delay)
## [1] 9317
summary(weather_delay$WEATHER_DELAY)
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##    1.00   11.00   22.00   43.64   54.00  297.00
nrow(weather_delay)
## [1] 1371
summary(security_delay$SECURITY_DELAY)
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##    1.00    8.00   14.00   17.84   21.00   81.00
nrow(security_delay)
## [1] 117

Summary

Our project investigated flight delays out of Boston Logan International Airport for domestic flights in 2015 to identify the most time-efficient trips for travelers from Boston. By summarizing and organizing the original dataset, we examined the relationships between flight delays out of Boston with the destination, airline, time of day, and also examined whether or not departing late guarantees a flight will also arrive late.

In 2015 flights from Boston departed to 62 airports within the United States. The most popular airports by flight volume were DCA, LGA, and ORD with each airport receiving over 6,500 flights from Boston Logan. The summer months had the most flights, and Saturdays and Sundays were the least popular days to fly.

We then explored what months, time of day, and airlines experienced the most delays. February had the highest delay time on average in minutes for the year of 2015, this was likely caused by the fact that February was the snowiest month on record in Boston since 1891. Also, flights typically experienced longer departure and arrival delays in the morning as compared to the evening. Finally Jet Blue had the most number of flights departing early.

We dug in several different ways to explore the relationship between departing late and arriving late. Of the 39,000 flights that departed late, 12,000 still managed to arrive at their destination airport on time. We found that in general, the longer a flight is the more likely it would be to arrive on time despite experiencing a flight delay.